Co-founder and CEO at 10 Senses
Currently, Artificial Intelligence (AI) and Large Language Models (LLMs), like ChatGPT, Mistral, or Gemini, are widely popular topics that never get off the posters.
All around the world, they spark discussions not only about the technology behind these models but also about the ethics and privacy of data. The latter ones are particularly vivid with the upcoming EU AI Act, which is a piece of legislation trying to regulate AI systems.
It classifies such systems into four categories:
- completely prohibiting those that fall into the “unacceptable risk” category,
- imposing certain requirements for those classified as “high-risk” models,
- demanding transparency from those that fall into the “limited risk” bucket.
The European legislation seems more than needed as we are facing the exponential growth of LLMs and each of them is upgraded with new cutting-edge functionalities on a regular basis.
Nevertheless, what average Joe doesn’t see is the tug-of-war that goes between continuous quality upgrades to these models and improving the user experience (UX).
Although the topic of UX in AI is getting more and more attention, it still doesn’t get as much spotlight as it should. The focus is still more on making LLMs smarter, but making them more enjoyable is lost in the process.
This article is the second part of the short series on how to improve UX in AI. The first part, where we covered the basics of UX in LLMs and discussed how improving searching capabilities can help with it, is here.
Now, we will look closer at a second lever that can diametrically transform the user experience, and this is the implementation of actions with one click.
What is the current issue with UX in LLMs?
As we already mentioned in the first part of the series, the first prompts that we use in the LLM chatbot usually don’t satisfy our needs. As a result, we ask more questions, providing more details, specifics, or precise commands.
Nevertheless, if you use Large Language Models in your work, business, or school, such prompts are usually repetitive.
For example, you may often write the same specifics, like “I’m the [work position] from [company] selling [product or service]. Write a persuasive follow-up email / summarization / translation to [name] from [company name] who is interested in our services / products during [event where you met]. Emphasize the benefits, such as [list benefits]. Keep it under [number of words].”.
Consequently, you may end up in a situation where you give an AI chatbot the same tasks, only changing the details depending on the recipients, audience, processes in the company, or assignments.
What if instead you could predefine certain basics, like:
- the number of words you need,
- the work position,
- the process name,
- the product or service you are selling,
- the type of document that you need,
and be able to access them with one click?
Improving UX in AI with one-click actions
Speeding up daily tasks
If you had such a capability in the Large Language Models, you would be able to not only speed up your daily work but also gain a lot of time for more productive tasks.
Instead of typing the same prompts again and again, you would have a list of predefined actions to choose from.
Standardized, manageable process
Such a solution could also streamline the process of prompting in companies that use LLMs in their daily operations.
One-click actions would allow for the standardization of prompts and would facilitate managing the process and onboarding new people in the team. Moreover, it would also be beneficial when employees leave the organization, as knowledge would stay in the company.
Truth be told, one-click actions would probably require a little bit of time and focus at the beginning. You would have to agree within the team or come up with which prompts need standardization and what the criteria are. Luckily, the initial effort would optimize processes quickly and pay off in a very short time.
Examples of one-click actions
Being equipped with such a capability, you could define tasks such as:
- Do a summary of,
- Translate to,
- Write e-mails,
- Write articles about,
- Write notes,
- Behave as if,
and many more.
Once you come up with all the phrases you need, you will be able to enclose them in a set of keywords you would define yourselves or within teams. Therefore, the final product would be a standardized list of actions with predefined criteria accessible with the click of a mouse.
Summing up, UX in Large Language Models should be discussed more often. Although we tend to focus on technology and legislation, these are the end users that are the key stakeholders when it comes to AI chatbots usage, so user experience should be prioritized.
Equipping them with the possibility to add a predefined toolkit of actions instead of making them copy or write the same prompts each and every day would significantly improve their experience, standardize the prompting in companies, optimize process flows, and increase the user experience in AI.
Talk to our expert
Are you looking for expert skills for your next data project?
Or maybe you need seasoned data scientists to extract value from data?
Fill out the contact form and we will respond as soon as possible.